Accept digits that are convertible to int for backwards compatibility

python · matthieucan · Jul 20, 2025 · Jul 20, 2025 · Jul 20, 2025 · Jul 20, 2025
commit f12e424b92a0722258b85c6553ad9e94e218155c
@@ -2398,21 +2398,31 @@ def get_section(value):
     The caller should already have dealt with leading CFWS.
 
     """
-    def is_ascii_digit(d):
-        # We don't use str.isdigit because only ASCII digits are allowed.
-        return '0' <= d <= '9'
+    def is_accepted_digit(d):
+        # While only ASCII digits are allowed by the RFC, we accept any digit
+        # that can be converted to an int for backwards compatibility purposes.
+        # We don't use str.isdigit() as some Unicode digits are not convertible
+        # (e.g. superscript digits).
+        try:
+            int(d)
+            return True
+        except ValueError:
+            return False
 
     section = Section()
     if not value or value[0] != '*':
         raise errors.HeaderParseError("Expected section but found {}".format(
                                         value))
     section.append(ValueTerminal('*', 'section-marker'))
     value = value[1:]
-    if not value or not is_ascii_digit(value[0]):
+    if not value or not is_accepted_digit(value[0]):
         raise errors.HeaderParseError("Expected section number but "
                                       "found {}".format(value))
     digits = ''
-    while value and is_ascii_digit(value[0]):
+    while value and is_accepted_digit(value[0]):
+        if not '0' <= value[0] <= '9':
+            section.defects.append(errors.InvalidHeaderDefect(
+                    "section number has a non-ASCII digit {}".format(value[0])))
         digits += value[0]
         value = value[1:]
     if digits[0] == '0' and digits != '0':

@@ -2983,13 +2983,20 @@ def mime_parameters_as_value(self,
             [('r', '"')],
             [errors.InvalidHeaderDefect]*2),
 
-        # gh-87112: Only ASCII digits can be section numbers.
-        'non_allowed_digits': (
+        # gh-87112: Only digits convertible to integers can be section numbers.
+        'non_accepted_digit': (
             'foo*0=bar; foo*²=baz',
             ' foo="bar"',
             'foo*0=bar; foo*²=baz',
             [('foo', 'bar')],
             [errors.InvalidHeaderDefect]),
+
+        'non_ascii_digit_backwards_compatibility': (
+            'foo*0=bar; foo*߁=baz',  # NKO digit '1'
+            ' foo="barbaz"',
+            'foo*0=bar; foo*߁=baz',
+            [('foo', 'barbaz')],
+            [errors.InvalidHeaderDefect]),
     }
 
 @parameterize